Hacker Newsnew | past | comments | ask | show | jobs | submit | jerb's commentslogin

Yes. Productivity tools make programmer time more valuable, not less. This is basic economics. You’re now able to generate more value per hour than before.

(Or if you’re being paid to waste time, maybe consider coding in assembly?)

So don’t be afraid. Learn to use the tools. They’re not magic, so stop expecting that. It’s like anything else, good at some things and not others.


But it’s clearly statistically much safer (https://www.tesla.com/VehicleSafetyReport) 7 million miles before an accident w FSD vs. 1 million when disengaged. I agree I didn’t like the feel of FSD either, but the numbers speak for themselves.


Teslas numbers have biases in them which paint a wrong picture:

https://www.forbes.com/sites/bradtempleton/2023/04/26/tesla-...

They compare incomparable data(city miles vs highway miles), autopilot is also mostly used on higways which is not where most accidents happen.


Thanks, I’ve never heard this and it’s quite profound. It’s always bothered me that there even is a top speed, and further that mass becomes infinite as it’s approached. But “speed of causality” makes these less strange.


Why does speed of causality make it any better? It’s still an arbitrary limit that’s even more abstract and intangible than speed.


If the universe’s causal mechanisms were infinitely fast, the entire history of the universe would play out instantly in zero time, and we’d skip straight to the heat death of the universe.

The fact that time even exists is implied by / a result of causal actions having some finite propagation time.


> causal actions having some finite propagation time.

I think I know what you're getting at, but somehow the phrasing bothers me, as if there is meta-time or as if cause and effect have time between them... for the photon at light speed, time isn't passing, it's emitted and then zero "time" later it hits something very far away.

It's more like we somehow need to think of cause and effect chains that have orderings without time.

I wonder if future generations will ever look back and casually quip something about "well they believed X existed, that was their problem, it all makes intuitive sense if you just..."


Yes, the basic idea is that photons do not 'experience' time. They 'experience' creation, all points along their path, and absorption 'simultaneously'.

However, you have to be careful with terminology. There is no inertial frame co-moving with the photon. All we can say is: as a massive particle gets faster relative to an observer's frame, the time it experiences relative to the observer's frame becomes shorter, and in the limit, as it approaches the speed of light (but never reaches c), the experienced relative time approaches zero (but never reaches 0).

This is well explained by Don Lincoln on the Fermilab YT channel:

Do photons experience time?

https://www.youtube.com/watch?v=6Zspu7ziA8Y


This is not true. You can easily measure TOF of a photon using an APD in Geiger mode or other detectors easily purchasable in 2024.

So what exactly are you talking about?


> for the photon

he’s talking about the photons frame of reference


It’s still not true.


That's really profound - thanks for sharing that insight. I've never looked at it that way before.


That doesn’t work that way.

Imagine a universe simulated in a computer with “ticks” where the entire state is updated.

It would be different to ours, but it would work just fine.


While what you say is true, it seems to me like what you're describing isn't quite the same definition of causality being infinitely fast.

The fact that causal effects happen in the next tick means some minimal time has passed. So in your definition causality can cover the entirety of a finite universe in an incredibly short amount of time (one tick). But it seems like that's not the same as covering the entirety of a finite universe in zero time. In that case, every result would happen within the same tick as its cause.


There's more than one way to "implement a universe", but one hypothetical way to run a simulation is to alternate the simulation of forces and effects/causes.

A simulation could use particles with attributes like position, velocity, and also a "sum of forces". Then each update has the following steps:

1. Reset the sum of forces for all particles to zero.

2. For all particles, add the contribution to the forces on it from all other particles.

3. For all particles, update the next particle position based on the collected forces.

In the above, there's no intermediate state between updates, everything moves to the next position synchronously from the perspective of in-universe observers. (The external simulation can update particles one-at-a-time, but this is not an "observable" inside the simulation.)


The cellular automata view of reality is entirely compatible with what I said.


Well because its decoupled from light itself or its properties.

If there’s some universal limit for causality itself, then light just happens to be the fastest thing among all the other things subject to causality.


But it is not a coincidence. Light — the EM field waves — propagates at causality speed because the EM field respects a particular property of the universe, the so-called gauge symmetry. That is intimately connected to the fact that the photobs has no mass.

Other similar particles, like the W and Z bosons, are manifestations of the weak field. Since that field breaks the symmetry, those particles have mass and move slower.

BTW, that symmetry breaking is the very same one that physicists talk about when we discuss the Higgs boson.


First person in this thread to make any sense… The level of physics knowledge has declined on this site.


I want to be wildly optimistic too, but I still see no evidence LLMs generate new knowledge. They always hew in-distribution.

Please correct me if I’m wrong


It's not new, but it's more usable, which makes new transactions and productions possible by lowering information costs.

That lower of costs is the ONLY basis for thinking AI is good for all. It's to the detriment of people previously managing the complexity manually through training and experience, but in favor of their customers who couldn't previously afford them.


They are really helpful to answer concrete questions, basically a replacement of manual web search and filtering, getting just exactly answers I was looking for. For example, one of my dialogue with chatgpt 4o was about boosting plant growth in a fish tank - you certainly can find a lot of web sites about it, but I simply described my aquatic environment and asked for a recipe - and the answers sound and well supported, a post dialogue double check of the sources help


That's the crucial point, though, you wouldn't have trusted it without checking the source after all. My experience with 4o is that 4 out of 5 answers I check are wildly incorrect or entirely made up, while with the rest, if I copy my prompt 1:1 into Google, I get the same correct answer pretty much verbatim in result #1. So I don't understand how so many people still see this as anything else but a waste of time, an intermediary step before going to an actual, credible source -- a step which in the best case is entirely unnecessary, in the worst case dangerously misleading. (From my non-representative survey among friends, the answer may be that barely anyone checks the validity of answers, and are simply unaware that they rely on answers from a system which, except for some lucky cases, gives them wrong ones.


This isn't snark - almost zero humans produce new knowledge.


When I say “new knowledge” I mean in the David Deutsch sense. Like the discovery of new physics which Sam mentioned.


T


Is the score of 28 comparable to the score of 29 here? https://www.kaggle.com/competitions/ai-mathematical-olympiad...


No. I would say it is more impressive than 50/50 there. (Source: I used to do math comps back in the day sorry it's not a great source)


IIUC the American Math Olympiad has 3 rounds. Wining the last one is almost a guaranty gold medal.

The link you posted has problems with a dificulty between the first and second round that are much easier.

I took a quik look at the recent list of problems in the first and second round. I expect this new AI to get a solid 50/50 points in this test.


Define “waste”. Then explain why your opinion should trump a free market. Then tell me how much coercion will be required to enforce it.


Problem with most chairs (including Aeron) is the bowl-shaped seat pan. It feels comfy / cradling when you first sit down, and that’s why people buy them. But as the minutes and hours go by, your pelvis is turning inward for lack of central support. Eventually you get back problems. Google Esther Gokhale for full explanation. An upwardly curved or at least flat seat pan (think old fashioned upholstered chairs or solid wood chairs) are more human designs.


> I often wonder what Minsky would think about the current generation of AI.

I suspect he'd react similiarly to Chomsky who in, a recent interview (MLST), was highly critical of LLMs as "not even a theory" (of what, i'm not sure... language aquisition? language production? maybe both)

Minksy was more broadly critical of NNs because it wasn't clear how difficult the problems they solved actually were. Until we had a better measure of that, saying "I got a NN to do X" is kind of meaningless. He elaborates in this excellent interview from 1990, beginning at 45:00: https://youtu.be/DrmnH0xkzQ8?t=2700


Scientific progress has slowed, but engineering progress is only just beginning. For instance, the electronic transport chain of respiration/photosynthesis, is a series of quantum tunnels. Man has barely scratched the surface of quantum-level control which nature already exhibits.


The number of transistors produced in 2021 is a bit mind boggling if you do the math. These are nanotechnology produced at a rate by humans that rivals any large physics number. By my calculation, Apple alone via TSMC produces something like 1x10^18 transistors per year. Add up Intel, Samsung and the rest of TSMC's customers and the number goes far higher. We are so used to it, we don't boggle at the concept any more but we should.


Consider applying for YC's Winter 2026 batch! Applications are open till Nov 10

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: